Twelve repositories for drawing it, building it and giving people somewhere to ask about it

Also turns text into diagrams and is far more widely supported, because GitHub, GitLab and most note apps render it without any install, but its layout is less able to untangle a dense architecture picture.

The oldest text-to-diagram tool of the three, covering more diagram types than either, but it needs Java installed and its output looks noticeably more dated.

The editor behind diagrams.net, and the opposite approach: you drag shapes with a mouse, which is quicker for one picture and much worse for keeping fifty of them current.

The thing that actually draws the map in a browser and reads the style file this editor writes, so it is the other half of the same job rather than an alternative.

Supplies the map data and a set of ready-made styles, which overlaps the styling part but through files you edit by hand rather than a window.

A smaller, newer project producing both the data and the cartography in one bundle, aimed at people who want a finished look rather than a tool for making one.

The closest match in kind: unstyled accessible React components covering much the same list, with a smaller following and a slightly lower-level way of composing them.

The same idea of behaviour without appearance, from the Tailwind CSS team and sized to pair with it, but it covers far fewer components.

Edition 34 covered this: the same job for SolidJS instead of React, so it is the answer for anyone not using React.

The field's first choice and the direct comparison: the same validation style and automatic documentation, far more widely known, but it leaves the database, settings and background-work decisions to you.

The small layer both of the above are built on, so choosing it means writing more yourself and carrying much less; renamed today from encode/starlette.

A third option with the same shape and a similar batteries-included view, smaller again and with a correspondingly smaller community.

The same job with tests written in Python rather than JavaScript, and a live web page showing the run as it happens, but it needs more machines to generate the same load.

A long-established alternative with very detailed reports, aimed at larger teams, written for the Java world and heavier to set up.

One tiny downloaded file that hammers a single web address and prints the timings, which covers the simplest case in one command and nothing beyond it.

The project that invented this way of writing requests and still the better-known name, with more features and plugins, but its last code landed on 17 December 2024 and nothing on its page says so.

Takes the other route to the same comfort: it is a thin front end over curl itself, so every curl feature is still there, but its own newest code is from 7 December 2025.

Edition 36 covered this: requests written in a plain text file with assertions, which is better for repeatable checks and worse for a quick look.

Edition 35 covered this: a much larger command set for CSV specifically, faster on the common jobs, but no JSON reshaping and no query language of its own.

Edition 36 covered this: the same CSV work with a strong emphasis on exploring and drawing in the terminal, which Miller does not do at all, and a smaller command set.

Takes the other approach entirely and lets you write real SQL against a delimited file, which is easier if you already know SQL and slower on large files.

Pure Python with nothing compiled, so it installs anywhere without a compiler and carries a permissive BSD licence, but it is much slower on large files and cannot render a page to an image.

Edition 35 covered this: also fast and also built on a C++ engine, with an MPL-2.0 licence that has no commercial condition, but it does not render pages and does not extract reading order.

A community fork focused narrowly on getting text and its layout out of a PDF, with an MIT licence and no editing, drawing or rendering at all.

Does the same job for JavaScript alone, listing every node_modules folder with its size, which is tidier for a machine that only holds JavaScript work and useless for anything else.

The Rust-only equivalent, and more precise there because it can delete only build output older than a given date, but its own page states that the project is currently unmaintained and looking for a maintainer.

Edition 36 covered this: it shows what filled a disk and can delete anything at all, which is more powerful and much less safe, because it has no idea which folders are rebuildable.

The closest match in spirit and simpler, showing a Markdown file as terminal slides and also able to run code blocks, but with fewer themes and no image support at all.

Edition 21 covered this: the same Markdown-to-slides idea but producing an HTML deck, a PDF or a PowerPoint file, which is what you want when the slides have to be sent to somebody.

The long-standing browser answer, far richer in layout, transitions and interactivity, and correspondingly more work to set up and not usable over a plain remote connection.

The official tagger of the MusicBrainz database and the strongest alternative: better at identifying unknown files, including by acoustic fingerprint, but it is a drag-files-in window rather than a grid, so bulk column editing is awkward.

Edition 36 covered this: the same tidying job done from the command line and as a catalogue you can query, which is far more powerful for automation and has no grid to look at.

A mature KDE tag editor covering many formats and the third real option on Linux, but its GitHub copy is a mirror of invent.kde.org, so the star count shown there means nothing about the project's size.

The closest direct alternative: the same forum job, written in Node.js, lighter to run on a small machine, but with a smaller plugin ecosystem and nothing like Discourse's trust-level system.

A PHP forum that is far simpler to install on ordinary shared hosting, and correspondingly less capable: extensions are fewer and several basics are add-ons.

Edition 34 covered this: shaped as a question-and-answer site rather than a conversation, which is better for a help centre and worse for a community, and much smaller to run.